SIGNAL RECONSTRUCTION FROM MULTIPLE CORRELATIONS - FREQUENCY-DOMAIN AND TIME-DOMAIN APPROACHES - COMMENT

Giannakis [J. Opt. Soc. Am. A 6, 682 (1989)] presented several approaches to reconstruction of one-dimensional deterministic sampled signals from their multiple correlations. I discuss the time-domain approach of Giannakis for reconstruction of infinite-duration signals. For model selection, rank determination of a certain matrix is needed. I give a counterexample, showing that the rank of such a matrix is not necessarily what has been claimed by Giannakis. For parameter determination, a certain other matrix is assumed to be of full rank. I give an example showing that the full-rank assumption is not always true.